Kalman filter based Variable Bit Rate video frame size prediction

Dynamic bandwidth allocation based on network traffic prediction is an important area of multimedia networking. In this paper, we present a frame size traffic prediction for MPEG-4 (moving picture expert group) VBR (variable bit rate) videos. Accurate prediction of VBR video traffic can be used to improve the network utilization while supporting the requirement of VBR video. In this paper we propose a Kalman filter based estimation of video frame size and compare its performance with other time domain prediction methods (normalized NLMS). Our proposed method uses past values for estimation and therefore can be used for online traffic measurement. Simulation results show that our proposed algorithm is effective in predicting the future traffic.